Hello Dino, The List,

That’s pretty cool to see activity around the document however I am not sure 
the proposed
changes are really addressing the structural problem of the document.

The current document is a mix between data-plane, control-plane, and operation 
questions.

The chairs proposition of re-balancing the text between 30bis, 33bis and an OAM 
documents
is great. It would allow people to go directly to the point they are interested 
in. 

1. What goes on the wire: 30bis
2. Signalling procedures: 33bis
3. Implementation details, management, and troubleshooting: OAM.

So it would mean that in 30bis it would just be all what is strictly needed to 
allow
inter-operability between xTRs, so at the end only packet format and how to 
understand
fields should be there. In this case, we can abstract the xTR as just a 
database that
stores mapping, how mapping are managed in this database is an implementation
question that is independent of the protocol itself.

For 33bis, it would just be the format of signalling messages and how to 
interpret
them, when a signal is expected to be triggered.

Finally, in OAM it would be how to implement and manage a LISP system that is
constituted of the LISP control-plane proposed in 33bis and the LISP data-plane 
in
30bis.

To say it clearly: remove from 30bis and 33bis all what is just the reflect of 
one
implementation. Normally these two document should have only what is strictly
necessary for people to implement (the way THEY want) a system that would
Inter-operate with the others. 

If we look at OpenLISP and its control-plane and the deployment of LISP-Lab
that we use in production daily, we can see that the data plane and control 
plane
have been implemented independently (and by different teams and even
companies) and what we can say is that a large fraction of the text in 60bis
has not been used at all for implementing the data-plane, while, on the contrary
we had to massively read/use text from 30bis to be able to implement the
control-plane. Finally, people that deployed LISP-Lab had to take content
from both 30bis and 33bis to be able to have a working environment. That
demonstrates that the separation is not done properly as normally people 
in charge of deploying a network should not have to read the data-plane
specs, or people implementing a control-plane should not have to read 
data-plane specifications.

I think the proposition of moving text that the chairs proposed is very
reasonable and greatly improve the quality of the specifications and therefore
reduce the risk of misinterpretation and bugs while implementing the protocolS


Cheers,

Damien Saucez 

> On 4 Jan 2018, at 18:00, Dino Farinacci <[email protected]> wrote:
> 
> Is the working group okay with me submitting these changes? This was the 
> latest update from email before the year ended. I have made most of the 
> changes that Luigi suggested or requested.
> 
> Dino
> 
> <rfcdiff-rfc6830bis.html>_______________________________________________
> lisp mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/lisp

_______________________________________________
lisp mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/lisp

Reply via email to